AI and Networksedit
Primary research topic for Qiao Xinbao, covering AI systems under networked data and communication constraints.
AI and Networks is the primary research topic currently emphasized in Xinbao Qiao's wiki. The term is used here in a deliberately broad but bounded sense: it covers AI methods for networked settings, and learning algorithms whose behavior depends on communication, decentralization, edge devices, institutional data silos, or cross-party evaluation.1
Introductionedit
In this wiki, AI and Networks is not a separate application label but the organizing frame for research in which learning is shaped by where data live, how information moves, and which parties can evaluate a model. The topic therefore includes decentralized learning, distributed computing, data pruning, collaborative evaluation, and synthetic-data verification under siloed access.
Role in this wikiedit
This page is the top-level hub for research in which model performance is shaped by where data live and how information moves. It links Qiao's background in communication engineering with later work on distributed learning, data silos, collaborative evaluation, distributed Wasserstein barycenters, and data pruning for decentralized training. The page also explains why several apparently separate projects are grouped together: they all treat communication, locality, or infrastructure as part of the learning problem, not merely as deployment details.
Current doctoral focusedit
In the CUHK doctoral stage, Qiao's recent focus within this topic is distributed computation for Wasserstein barycenters. The problem is a natural fit for AI and networks: each participant can hold a local empirical distribution, while the system needs a collective distributional reference for comparison, verification, or control. This places the emphasis on algorithms that respect communication and data-access constraints, not only on centralized statistical objectives.
Publicationsedit
| Paper | Venue/status |
|---|---|
| When Sample Selection Bias Precipitates Model Collapse | ICML 2026, 6-11 July 2026, Seoul. |
Connection to Qiao's workedit
The ICML 2026 paper When Sample Selection Bias Precipitates Model Collapse belongs here because it studies collaborative verification under siloed access. The current Wasserstein-barycenter focus continues that thread by treating a reference distribution as something that may need to be computed across a network rather than assumed to exist centrally. Earlier work in machine unlearning contributes the same systems instinct: algorithms are evaluated not only by accuracy, but also by latency, communication, and the cost of changing data after training.
See alsoedit
- Distributed Learning
- Data Silos
- Collaborative Evaluation
- Distributed Wasserstein Barycenter
- Data Centric ML
- The Chinese University of Hong Kong
Footnotesedit
-
The topic label follows CUHK IE's official department description, which frames information engineering around information generation, communication, storage, and processing in real-world applications; the ICML 2026 timing in the publications table follows the official ICML 2026 conference page. ↩